ARTICLE
A Multiscale Study of High Performance
Double-Walled Nanotube!Polymer
Fibers
Mohammad Naraghi,†,§ Tobin Filleter,†,§ Alexander Moravsky,‡ Mark Locascio,† Raouf O. Loutfy,‡ and
Horacio D. Espinosa†,*
†
Department of Mechanical Engineering, Northwestern University, 2145 Sheridan Road, Evanston, Illinois 60208, United States, and ‡MER Corporation, 7960 South Kolb
Road, Tucson, Arizona 85706, United States. §These authors contributed equally to this work.
T
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
he study of natural materials pro-
vides great insight into the hierarchi-
cal design of structures to achieve
Downloaded via SHENZHEN UNIV on September 4, 2020 at 14:42:01 (UTC).
remarkable mechanical performance with
both high strength and robustness. The su-
perior mechanical performance found in na-
ture is in contrast to common engineering
materials, such as ceramics and metals, in
which there is a trade-off between strength
and toughness.9,13!17 Hierarchical designs
in nature, such as in bones and proteins,
span a wide range of length scales in which
the materials are known to be significantly
robust and insensitive to flaws with tough-
ening mechanisms present across nano-
scale, mesoscale, and macroscopic
lengths.3,16,18!21 The hierarchy in nature has
motivated attempts to mimic natural struc-
tures toward the development of high
strength/toughness materials.18,22!27
One promising building block to synthe-
size biomimetic hierarchical structures is
carbon nanotubes (CNT), which in their pris-
tine form offer significant strength and
modulus, in the order of 100 GPa and 1
TPa, respectively.28!30 In addition, due to
their high thermal stability and electrical
properties, they can add thermal and elec-
trical functionalities to the constituting
structures.1,31!33
However, recent attempts to incorpo-
rate CNTs into hierarchical structures point
to the inefficient integrity between the
neighboring CNTs and/or the CNTs and the
other constituents of the macrostructures.
As a result, the superior mechanical proper-
ties of CNTs are only marginally utilized.
Windle et al. have devised and imple-
mented an in situ chemical vapor deposi-
tion (CVD) spinning method to obtain con-
tinuous CNT yarns from CVD grown CNT
aerogels.1,2,34 The stretching of the low den-
www.acsnano.org
ABSTRACT The superior mechanical behavior of carbon nanotubes (CNT) and their electrical and thermal
functionalities has motivated researchers to exploit them as building blocks to develop advanced materials. Here,
we demonstrate high performance double-walled nanotube (DWNT)!polymer composite yarns formed by
twisting and stretching of ribbons of randomly oriented bundles of DWNTs thinly coated with polymeric organic
compounds. A multiscale in situ scanning electron microscopy experimental approach was implemented to
investigate the mechanical performance of yarns and isolated DWNT bundles with and without polymer coatings.
DWNT!polymer yarns exhibited significant ductility of "20%, with energy-to-failure of as high as "100 J g!1,
superior to previously reported CNT-based yarns. The enhanced ductility is not at the expense of strength, as yarns
exhibited strength as high as "1.4 GPa. In addition, the significance of twisting on the densification of yarns
and corresponding enhancement in the lateral interactions between bundles is identified. Experiments at
nanometer and macroscopic length scales on DWNT!polymer yarns and bundles further enabled quantification
of energy dissipation/storage mechanisms in the yarns during axial deformations. We demonstrate that while
isolated DWNT bundles are capable of storing/dissipating up to "500 J g!1 at failure, unoptimal load transfer
between individual bundles prevents the stress build up in the yarns required for considerable energy storage at
the bundle level. By contrast, through polymer lateral interactions, a much better performance is obtained with
the majority of energy dissipated at failure being contributed by the interactions between the polymer coating and
the DWNTs as compared to the direct van der Waals interactions between bundles.
KEYWORDS: carbon nanotubes bundles · carbon nanotubes yarns · energy-to-failure ·
nanocomposites · in situ mechanical testing
sity aerogels of CNTs with low entangle-
ment generated inside the CVD reactor al-
lowed for significant alignment of the CNTs
leading to high mechanical performance of
the yarns. Using in situ CVD spinning tech-
nology, the Windle group was able to fabri-
cate CNT yarns with specific strengths as
high as "9.5 Ntex!1 at small gage lengths
of "1 mm. Defects resulted in a substantial
reduction of specific strength to "1 Ntex!1 *Address correspondence to
espinosa@northwestern.edu.
at 20 mm gage length and energy-to-
failures to "30!65 J g!1.2,9,35 These values Received for review June 21, 2010
are comparable to the properties of high and accepted October 19, 2010.
performance fibers such as Kevlar as listed
Published online October 26, 2010.
in ref 2. However, the yarns fabricated by 10.1021/nn101404u
this method are not highly uniform in diam-
eter, which can potentially induce stress © 2010 American Chemical Society
VOL. 4 ▪ NO. 11 ▪ 6463–6476 ▪ 2010 6463
ARTICLE concentrations and cause premature failure.10 In addi-
tion, it has been suggested that the relatively large di-
ameter of the CNTs and their consequent collapsed
structure yields enhanced mechanical performance;
however, further understanding of the role of van der
Waals interactions is still needed to elucidate this
enhancement.2,34!36
Another approach toward developing high perfor-
mance structures from CNTs was proposed by Baugh-
man et al.37 which included extracting CNT yarns from
forests of CNTs by the introduction of twist to the CNTs.
A high degree of alignment was achieved by this
method. The strength and energy-to-failure of their
samples was ca. 0.3!0.6 Ntex!1 and 14!20 J g!1,
respectively.3,38 This process was further enhanced by
Liu et al., in which yarn spinning was performed by a
combination of twisting and solvent assisted shrink-
ing of the CNTs from CNT forests. It was demon-
strated that shrinking and twisting enhances the me-
chanical performance of the yarns potentially due
to the increased interaction between CNTs.39 More-
over, Baughman et al.3 have infiltrated the spun
yarns with polymers such as polyvinyl alcohol (PVA)
to enhance the interactions between CNTs and the
mechanical performance of the yarns. These at-
tempts have led to slightly stronger yarns at the
cost of a reduction in ductility and consequently
lower energy-to-failure. Other approaches such as
drawing yarns from CNTs solutions and gels have
also been introduced, through which CNT-based
composites with high concentration of CNTs are
fabricated.1,4,6 These methods were primarily imple-
mented to allow for more interactions between CNTs
through polymer molecule intermediaries, poten-
tially increasing the yarns toughness and strength.
Despite the significant improvements in recent years
toward developing CNT-based yarns, these examples
all point to the fact that mechanical properties of
these structures, such as their strength and energy-
to-failure, remain significantly lower than those ex-
hibited by CNTs.
While the most significant contributing factors in
the relatively poor mechanical performance of CNT-
based macrostructures have been identified as inferior
hierarchical designs and low load transfer between
CNTs, a more detailed fundamental understanding of
load bearing mechanisms and energy storage/dissipa-
tion mechanisms is necessary to bridge the gap be-
tween the mechanical properties of these macroscale
samples and their nanoscale constituents. A number of
fundamental studies on individual CNTs and CNT
bundles have begun to develop this desired fundamen-
tal understanding. Ruoff et al. performed in situ scan-
ning electron miscroscopy (SEM) tensile tests of both
multiwalled nanotubes (MWNTs) and single-walled
nanotube (SWNT) bundles, experimentally demonstrat-
ing the exceptional mechanical properties of indi-
6464 VOL. 4 ▪ NO. 11 ▪ NARAGHI ET AL.
vidual tubes, as well as identifying weak shear interac-
tions between adjacent MWNT shells.40,41 Through in
situ transmission electron microscopy (TEM) testing, our
group has experimentally measured, for the first time,
CNT modulus and strength consistent with quantum
mechanical predictions.29 Moreover, the advantages of
covalently cross-linking inner shells of MWNTs as well as
the SWNT within bundles to improve load transfer be-
tween shells have also been demonstrated.28,29,42
In the current study, in order to develop a funda-
mental understanding of the effect of interactions be-
tween CNTs on the mechanical performance of CNT-
based yarns, we have carried out a multiscale
experimental approach to quantify the effect of shear
interactions via polymer cross-linking of CNTs. Here,
CNT-based yarns were fabricated from chemical vapor
deposition (CVD) grown double-walled nanotube
(DWNT) mats. The mats were composed of bundled
DWNTs, coated in situ with approximately a monolayer
of organic polymer during the CVD process. DWNTs in
particular have the advantage of a higher density of
load bearing graphitic shells as compared to larger
MWNTs. The organic layer has been found to substan-
tially enhance the interactions between CNTs, thus im-
proving the mechanical performance of the yarns. Fol-
lowing thermal treatments, the polymer coating was
removed, considerably altering the interactions be-
tween adjacent CNTs and providing a comparison in
mechanical properties. We have adopted a unique ap-
proach in which we study both the macroscopic yarns
as well as their nanoscale constituents, individual DWNT
bundles, and we use the results of the latter to inter-
pret the former. This approach allows an unambiguous
comparison of the material properties on multiple
length scales within the hierarchical network, leading
to fundamental insights into the relative effectiveness
of different load transfer mechanisms in CNT-based ma-
terials. Our multiscale experimental approach is com-
posed of in situ SEM nanoscale experiments on indi-
vidual bundles and in situ SEM tension tests on
macroscale yarns, in which the mechanical perfor-
mance of the bundles and yarns, their modes of fail-
ure, and the correspondence between the two scales
are identified. Through our multiscale approach, we
demonstrate that an enhancement in lateral interac-
tions between CNT bundles as a result of polymer cross-
linking has significantly improved the mechanical per-
formance of the DWNT-based yarns.
RESULTS
DWNT yarns were fabricated by spinning DWNT
mats produced by MER Corp. in a CVD process by us-
ing a floating catalyst system based on ferrocene as the
catalyst precursor and ethyl alcohol as a hydrocarbon
feedstock. The mats are composed of an interwoven
network of DWNT bundles. Figure 1 shows the hierar-
chy of the yarn structure down to individual DWNT
www.acsnano.org

Figure 1. Hierarchy of DWNT mat material. (a) Individual DWNT bundle. Scale bar: 10 nm. Insets: (top) crossection of indi-
vidual bundle; (bottom) nanodiffraction pattern of an individual bundle. (b) TEM image of interconnected DWNT bundles in-
cluding Fe particles. Scale bar: 100 nm. (c) SEM image of DWNT mat. (d) SEM image of a DWNT yarn.
bundles. Each bundle within the mat is composed of
up to tens of hexagonally close-packed DWNTs. These
DWNT building blocks have an outer diameter of "2.2
nm, determined from Raman spectra breathing modes
(not shown), and a spacing of "2.5 nm between adja-
cent DWNTs within bundles, measured from high reso-
lution TEM images. Nanodiffraction patterns of indi-
vidual bundles reveal variation in the orientation of the
CNTs lattice with respect to the electron beam in TEM
resulting in continuous bands in the diffraction patterns
(see Figure 1a). This can either indicate a variation in
chiral angle for CNTs within the bundles or misalign-
ment with respect to the bundle axis.43
In combination with other analytical techniques
such as FTIR, the thermogravimetric analysis (TGA)
of the as-produced DWNT mats in air and argon (Fig-
ure 2) enables identification and quantitative deter-
mination of the contents of the mat components
that include DWNTs, organic compounds, iron, amor-
phous and graphitized carbon. When organic com-
pounds are volatized by heating (Figure 2), they turn
into a multicomponent gas mixture that is domi-
nated by the molecules of substituted acrylic acid es-
ters with the general formula R=CHACR==COOR,
where R, R= and R== stand for hydrogen, alkyl, alke-
nyl, and aryl groups in various combinations, as
demonstrated by mass-spectrometric (MS) analysis.
Figure 2. Heat treatment to remove polymer coating. TGA
spectra in argon and air at 20 °C min!1 demonstrating the re-
moval of the majority of the polymer coating in argon at
600 °C.
www.acsnano.org
ARTICLE
Organic compounds can also be extracted from
DWNT mats by applying a set of common solvents
to the mat. FTIR and HPLC/UV analyses confirm the
dominance of molecules of substituted acrylates in
these extracts. Polyacrylates undergo scission into
monomers upon thermal decomposition and disso-
lution. The polyacrylate layer is formed in the CVD
reactor from the same acrylate monomers identified
above. The gas phase in the nanotube synthesis con-
sists of hydrogen, ethylene, acetylene, carbon mon-
oxide, methane, and water vapor mixture, which is
also well suited for the synthesis of acrylic acid de-
rivatives. The acrylate monomer synthesis takes
place in the moderately hot zone of the reactor after
the nanotube synthesis is completed and is
likely catalyzed by the surface of nanotubes, where
the polymerization is also occurring and the result-
ing polyacrylate layer eventually resides. Therefore,
based on the data above and electron microscopy
observations it is speculated that the organic com-
pounds in the as-produced DWNT mat exist as a thin
polymeric layer that coat the DWNT bundles, which
may also be present on the inner tubes.
As demonstrated in the TGA analysis, Figure 2, the
majority of the polymer coating can be vaporized and
removed at a temperature of 600 °C in argon (vertical
red line). Beyond 600!700 °C, little weight change is
observed in Ar. Moreover, TGA experiments in air dem-
onstrate a yarn-mat mass loss which reaches a local pla-
teau at ca. 250!400 °C (vertical blue line in Figure 2), at-
tributed to the burning of the amorphous carbon and
polymeric coatings. The upper limit in this temperature
range points to the onset of the burning of the SWNT.
These results specifically identify the heat treat-
ments required to remove the polymer coating with-
out modifying the sp2 bonding in DWNTs both in ar-
gon and in air.44,45 Further heating of the sample to
higher temperatures in air shows a rapid weight loss
starting at "450 °C and continuing to 700 °C, dur-
ing which DWNTs and iron catalysts are burned. The
TGA experiments in Ar point to the presence of
"14.5 wt % polymer coating in the yarns, while the
VOL. 4 ▪ NO. 11 ▪ 6463–6476 ▪ 2010 6465
ARTICLE TGA experiments in air reveal the presence of "35
wt % of Fe2O3 residues which translates to "24.5 wt
% of Fe in the as-fabricated mats.
On the basis of these findings, an argon!air heat
treatment was applied to the as-fabricated yarns and
mats (pristine samples) which contain the polymer
coatings to produce DWNT yarn and mats without the
cross-linking polymer coating (heat-treated samples) in
order to study the effect of polymer coating on the me-
chanical behavior of yarns and bundles. More details
of the heat treatment are presented in the experimen-
tal section.
Mechanical Behavior of DWNT Yarns. In-situ SEM mechani-
cal characterization of CNT-based yarns was carried out
using a tensile testing stage, allowing for simultaneous
measurement of the mechanical behavior of samples
and identification of mechanisms of deformation and
failure of the material.
Since both enhanced mechanical behavior and
lightweight are desired for most advanced structural
materials, here we have calculated the mechanical
properties of yarns in terms of specific properties such
as specific strength (ratio of yarn strength in units of
stress to density) and energy-to-failure per unit weight.
In addition, for comparisons with other structural mate-
rials and with individual DWNT bundles, we also report
our measurements of the yarns mechanical properties
based on estimates of the yarns cross section.
To calculate the yarns effective diameter, the cross-
sectional area of DWNTs in the yarns is determined from
the mass of CNTs within the sample and the density of
CNT bundles with a nominal density #bundle $ 1.575
g/cm3 as estimated in the Supporting Information. The
effective cross section of each yarn, by excluding the
volume of the pores between bundles, can be esti-
mated as
λ
Aeffective ) (1)
Fbundle
where % is the linear density of the CNTs in the yarns.
Since iron particles do not contribute to the load bear-
ing in the yarns, they are excluded in the effective area
measurements, as suggested by Motta et al.10 The effec-
tive diameter of the yarns is then calculated as the di-
ameter of a yarn with a circular cross section with an
area of Aeffective.
A sample of the specific stress-engineering strain
curve of a DWNT-based yarn obtained in this study
is shown in Figure 3a. In this case, the yarn was fab-
ricated from a 1/8 in. wide ribbon of a DWNT mat,
and the sample gage length was "1 cm. The sample
failed at a specific stress of "0.8 N/tex, which trans-
lates to "1.4 GPa. It also demonstrates a large duc-
tility of "20% engineering strain. The energy-to-
failure of this specific sample is "100 J g!1 of DWNTs
weight. Subsequent to the removal of the inherent
polymer coating, the specific strength slightly in-
6466 VOL. 4 ▪ NO. 11 ▪ NARAGHI ET AL.
creases to "1.0 N/tex (equivalent to "1.6 GPa),
while the ductility and energy-to-failure substan-
tially drop, Figure 3a. The results shown in Figure
3a are typical of the yarns made out of ribbons with
the same ribbon width (1/8 in.), as will be discussed
further in this section.
The fractography of the yarns points to an interest-
ing aspect of their morphology. From the SEM images
of the surfaces of the yarns, for instance Figure 1d, we
see that the yarns may resemble a continuous filament-
like structure fabricated through twisting a series of fila-
ments together,46 SEM images of the fracture surface in-
stead reveal their true structure. As shown in Figure 3b,
the filament-like structures observed on the surface
are wrinkled layers of the original ribbon used to fabri-
cate the yarn. This morphology is most likely a conse-
quence of successive buckling of the mat during spin-
ning which results in ridges similar to those formed in
thin cylinders after being twisted to the postbuckling
regime.47,48 It is important to note that the polymer
coating binds the wrinkled layers of the yarn, thus en-
hancing its structural stability and integrity. This feature
becomes apparent when monitoring the structure of
heat-treated yarns (no polymer coating) in which the
stacked wrinkled layers of the mat are partially unfolded
at several locations due to the loss of binding poly-
meric molecules, Figure 3c. It is however to be noted
that even in this case, van der Waals interactions be-
tween the neighboring mats brings a degree of struc-
tural stability to the yarns.
The in situ SEM experiments also reveal the mecha-
nisms of deformation of the yarns. As shown in Figure
3d!f, the axial deformation of the pristine yarns (before
heat treatment) is accompanied by aligning the ridges
of the buckled wrinkles with the loading direction. As
the yarn is stretched from an undeformed configuration
to an average engineering strain of "14% (cross head
displacement normalized by the initial yarn length), the
angle between the wrinkles’ ridges and the axis of the
yarn is reduced from "23° to "10.3°. In addition,
sample stretching is accompanied by the twisting, real-
ized for instance by tracking the motion of the par-
ticles on the surface of the yarn in Figure 3d!f. This
twist of the yarns during axial deformations demon-
strates the development of a torque during axial load-
ing. This mode of deformation is reminiscent of the me-
chanical behavior of the filament yarns under axial
load when the constitutive filaments have the same lay-
out (twisted in the same direction).46 The development
of this moment in the yarn upon axial loading can be
explained by the significantly higher stiffness of the
wrinkles in the direction of the ridges compared to the
lateral direction (direction of compressive stress during
buckling) along which due to buckling they have sub-
stantially lost their load bearing capability. As a result,
the axial force on the yarn is carried by the wrinkles in
the direction parallel to the ridges which makes a finite
www.acsnano.org
Figure 3. (a) Specific stress as a function of engineering strain in DWNT based yarns. (b) The post-mortem SEM images of
the fractured surface of the yarns clearly points to the buckled structure of the mat formed during spinning. (c) SEM image
ARTICLE
of the heat-treated yarns. The removal of the polymer coating reduces the interaction between layers of the mat in the yarns
causing partial unfolding. SEM images of a yarn during in situ SEM tension test at (d) zero load, (e) 6.8% eng. strain, and
(f) 14% eng. strain.

angle with the yarn axis ("23° to "10.3° in Figure 3),

thus a net torque is induced on the sample. A conse-
quence of the combined stretching and torsion on the
yarn is the enhancement of lateral confinement during
axial loading, and the very high Poisson ratio, noticed
by significant shrinkage of the yarn diameter ("13% re-
duction in diameter in "14% axial strain), shown in Fig-
ure 3d!f.
It is illustrative to estimate the material strain &material
by decomposing the total engineering strain &engineering
into two parts,49 as follows:
www.acsnano.org
where q0 ' tan((0), q ' (1 ) &engineering) · tan((), and *0
and * are the initial and current ridge angles. The first
term on the right-hand side of this equation represents
the yarn strain induced by the reorientation and tilting of
the ridges toward the loading direction, while the sec-
ond term represents the component of the true strain in
the axial direction of the yarn.
VOL. 4 ▪ NO. 11 ▪ 6463–6476 ▪ 2010 6467
ARTICLE

Figure 4. (a) Strength and (b) energy-to-failure as a function of diameter for the pristine and heat-treated yarns. (c) The
spinability of the yarns as a function of yarn effective diameter, for each yarn diameter the corresponding ribbon width is
shown on the plot. Postmortem SEM images of the cross section of the pristine yarns with an effective diameter of (d) 55,
(e) 38, and (e) 27 "m. Postmortem SEM images of the cross section of the (g) pristine and (h) heat-treated yarns, pointing to
the cooperative deformation of the bundles specific to the pristine yarns (g).

In this analysis, it is assumed that each yarn is com-
posed of oriented fibrils making an initial angle of
(0 $ 23° with the axis of the undeformed yarn,
Figure 3d. Hence, right before rupture we have * $
10.3°, &engineering $ 14%, and &material $ 6.3%. In other
words, out of 14% overall strain measured in the
sample, only about half will stretch the network of the
bundles, while the rest of it tends to align the ridges
with the loading direction and thus will transform the
macroscopic structure of the yarns to a more compact
structure (supported by the evident reduction in diam-
eter during axial loading augmenting to a Poisson’s ra-
tio of ca. !0.8 to !0.9). Therefore, we speculate that the
presence of the initial twist in the yarns has a positive
effect on the strength by compressing the yarn laterally
and increasing the interactions between bundles. More-
over, one may argue that the initial twist potentially in-
creases the strength beyond the strength of the un-
twisted yarn. This can be hypothesized by considering
the following two cases: a twisted yarn and yarn with no
twist. In both cases, near failure, the ridges will be nearly
parallel to the loading direction (consider Figure 3f for
the first case). However, the former will benefit from the
extra densification it will undergo during axial loading,
which is absent in the latter, leading to more
6468 VOL. 4 ▪ NO. 11 ▪ NARAGHI ET AL.
bundle!bundle interaction. However, this hypothesis
cannot be verified experimentally since the twisting
procedure is required to achieve strong yarns from the
initially very fragile mats. Therefore, it is believed that
the special morphology of the wrinkles of the yarns can
enhance both the ductility and strength of the yarns,
and as a consequence the energy-to-failure.
The specific strength and energy-to-failure of the
pristine yarns (with inherent polymer coating) as a func-
tion of effective diameter are shown in Figure 4 panels
a and b, respectively. Each point in these figures repre-
sents an average of a minimum of 3 measurements. The
graphs point to a monotonic decrease of the mechani-
cal properties of the yarns with effective diameter in-
crease. As the effective diameter increases from 27 to
53 +m, the specific strength and energy-to-failure de-
crease from 0.70 , 0.07 to 0.40 , 0.07 Ntex!1 (approxi-
mately from 1.4 , 0.1 to 0.8 , 0.1 GPa) and from 109
, 17 to 38 , 13 J g!1, respectively. The origin of the en-
hanced mechanical performance of the thinner yarns
is evident in the cross-sectional images of fractured
yarns, Figure 4d!f. As shown in the figure, the thinner
yarns have a more compact cross section (less porous)
especially in the core of the yarn due to the higher
twists imposed on narrower ribbons in the spinning
www.acsnano.org
 ARTICLE
Figure 5. Schematic representation of the deformation of pristine yarns and the cooperative load bearing through polymer
coating. Upon removal of the polymer, the network loses its ability to sustain large deformations. The cooperative deforma-
tion in the pristine yarns has a striking similarity with natural materials such as bone, in which a soft matrix (mainly com-
posed of collagen) is reinforced with a stiff inorganic component (minerals platelets) to attain both high toughness and
strength.12 The mechanical functionality of the collagen matrix and mineral platelets in bone are similar to the functional-
ities of the inherent polymer coating and CNTs in our developed material system.

process. In other words, narrower ribbons of mats al-
lows for a more uniform twist in different portions of
the mat, resulting in a more compact yarn throughout
the cross section. The higher amount of twist on nar-
rower ribbons is quantified by the number of twists per
unit length of the ribbons plotted as a function of the
ribbon width in Figure 4c. Wider ribbons (thicker yarns)
reach the critical force per unit width of the ribbons
(80 mN mm!1) at a smaller number of turns during spin-
ning. This is due to the higher shear strains that de-
velop during spinning on wider yarns at a given num-
ber of turns, which increases the stress in the yarn
during spinning. The higher capability of narrower rib-
bons to be spun without failure allows for the develop-
ment of a higher compactness, and thus it induces
stronger interactions between the CNT bundles inside
each yarn. This enhanced mechanical performance of
yarns with higher twist has been previously predicted.50
Figure 3a and Figure 4a,b also point to the signifi-
cant differences between the mechanical behavior of
pristine (with inherent polymer coating) and heat-
treated yarns. It is noteworthy to emphasize that the
main distinction between the yarns before and after
heat treatment is the amount of polymer content,
which is significantly lower in the latter. As shown in
Figure 3a, after the removal of polymer, the yarns ap-
pear to be significantly stiffer and more brittle. These
differences in the mechanical behavior of the yarns
point to the increased short-range interactions be-
tween bundles as a result of the removal of the poly-
www.acsnano.org
mer coating. The increased short-range interactions be-
tween bundles are motivated by the more compact
structure of heat-treated yarns evidenced by "15%
shrinkage of their diameter as a result of heat treat-
ment compared to the pristine ones. On the contrary,
in the pristine yarns, the polymer coatings between the
bundles increases the interbundle distances and acts
as a plasticizer facilitating the mutual sliding of the
bundles on each other especially at small deforma-
tions. As a result, the pristine yarns are significantly
more compliant.
The contribution of the polymer coating to the over-
all deformation of the yarn is clearly evident in the dif-
ferences observed between the fracture surfaces of the
pristine yarns (Figure 4g) and heat-treated yarns (Fig-
ure 4h). In the former, the DWNT bundles deform coop-
eratively, and they slide significantly on the neighbor-
ing bundles during the test, while in the latter individual
DWNT bundles pulled out resulting in a disordered
and unaligned morphology. In other words, the load
transfer between DWNT bundles via polymer molecule
intermediaries significantly expands the fracture pro-
cess zone in pristine yarns, allowing the material to dis-
sipate more energy before failure.
Despite the initial low stiffness of the pristine yarns,
their ultimate strength is only marginally affected by
the removal of the polymer coating, Figure 4a. This ob-
servation suggests that despite the initial weak load
transfer between bundles in the pristine yarns, where
the interbundle interactions are governed by polymer
VOL. 4 ▪ NO. 11 ▪ 6463–6476 ▪ 2010 6469
ARTICLE
Figure 6. In-situ SEM tensile testing setup used for mechani-
cal testing of DWNT bundles. (Inset) Isolated DWNT bundle
attached between Si tip and nanomanipulator tip. The force
is measured by the deflection of the Si tip.
molecules (initial low stiffness of pristine yarns), the
shear interactions between bundles improve as a re-
sult of yarn axial stretching. In other words, at low
strains, the interactions between bundles through poly-
meric molecules are low due to low elasticity of coiled
polymer macromolecules. However, at large deforma-
tions, the polymeric molecules get stretched to accom-
modate the deformation; therefore, they get stiffer en-
tropically, resulting in increased interactions between
bundles, which improves the load bearing in the axial
direction.
It is noteworthy here to mention the similarities be-
tween our CNT!polymer system with that of nature hi-
erarchical structures such as bone,12 in which a high
stiffness mineral and soft matrix are mixed to obtain en-
hanced mechanical performance. The analogy be-
tween the two systems can be realized by comparing
our material system, schematically shown in Figure 5,
with the collagen matrix of bone reinforced with
platelet-mineral shown in ref 12. There is a striking simi-
larity between the mechanical functionalities of the col-
lagen matrix (soft segment) and platelet minerals (hard
segment) in bone to inherent polymer coating and
CNTs in our developed material system, respectively.
In both cases, the cooperative deformation mecha-
nisms of the soft and hard segments allow for large ma-
terial ductility, while in the nanoscale the material con-
tains stiff elements. In both cases, the presence of the
soft segment has a crucial role in increasing the material
ductility.
Mechanical Behavior of DWNT Bundles. To identify the lim-
its of realizable strength and energy-to-failure of our
DWNT- based macroscale yarns, DWNT bundles with
and without polymer coating have been tested using
an in situ SEM tensile testing technique with a Si micro-
fabricated cantilever as the load sensor (see Figure 6).
For analyzing the mechanical properties of DWNT
bundles both the outer and inner shells of each DWNT
(consistent with the analysis of the DWNT yarns) on the
perimeter of the bundle was considered in calculating
the true stress. Stress calculations where based on
measured load and a cross-sectional area consisting of
6470 VOL. 4 ▪ NO. 11 ▪ NARAGHI ET AL.
Figure 7. Diameter dependence of the experimentally de-
termined elastic modulus for pristine DWNT bundles using
two different models for calculating stress. Model A consid-
ers all DWNTs within the bundle, whereas model B only con-
siders outer tubes.
DWNTs on the perimeter of the bundle as illustrated
by model B in Figure 7. The bundles were assumed to
be hexagonal close packed structures, approximately
circular in geometry, consistent with TEM imaging. In
nanoscale tests of individual bundles the load is applied
to continuous DWNTs on the perimeter of the bundle,
which span the full gauge region. The stress in the
bundles was calculated using the following equations:
F
σ) (3a)
Nπt(do + di)
where do and di are the outer and inner shell diam-
eters of individual DWNTs within a bundle, respec-
tively (2.2 and 1.6 nm). The number of load-bearing
DWNTs considered within a bundle is given by
π(dbundle - do)
N) (3b)
ds
where dbundle and ds are the diameter of the entire
DWNT bundle and the axes spacing of adjacent DWNTs
within a bundle (2.5 nm). Equations 3a and 3b are de-
rived in more detail in the Supporting Information. Such
a model, in which only the tubes around the perimeter
are considered in the cross-sectional area, is motivated
by previous studies of SWNT bundles.40 To further sup-
port this choice of model for calculating the true stress,
we have investigated the diameter dependence of the
elastic modulus of tested DWNT bundles. Figure 7
shows experimental results of the elastic modulus de-
termined using two different load bearing cross sec-
tional models for calculating stress. As expected, model
B (eq 3a and 3b) in which only the outer perimeter of
DWNTs are considered, is found to exhibit no variation
of modulus with diameter. This is consistent with iden-
www.acsnano.org

Figure 8. Stress!strain curves recorded for DWNT bundles
with (a) and without (b) heat treatment. The insets show
loading and unloading curves revealing plastic behavior for
pristine bundles and nearly elastic behavior for heat-treated
bundles. (c) Sword-in-sheath failure of heat-treated bundle.
Complete fracture (d) and telescopic failure (e) of pristine
bundles. All scale bars: 250 nm.
tical properties of all DWNTs within the bundle and
load being carried only on the outer tubes. In contrast,
when model A (in which all DWNTs within the bundle
are considered) is used to calculate the stress, the elas-
tic modulus is found to decrease with diameter, which
would not be expected if all DWNTs were carrying the
load and had equal mechanical properties. This con-
firms that the best physical description of the loading
conditions is that only the outer perimeter DWNTs are
carrying load and that model B best describes the true
stress acting on the load bearing DWNTs within the
bundle.
Figure 8 shows stress!strain curves recorded from
tensile tests performed on several individual bundles
with and without heat treatment. Using model B to cal-
culate true stress described earlier, the elastic modulus
of DWNT bundles was found to be similar for bundles
with and without polymer (ca. 200!250 GPa). Signifi-
cant variation in the strength and failure strain was
found for both heat-treated and pristine bundles. From
Figure 8 it is observed that the DWNT samples fail at
stresses ranging between "5 and 25 GPa. One distinct
difference in the mechanical response of pristine and
heat-treated bundles is that the pristine bundles exhibit
plastic yielding consistent with a stick!slip sliding
mechanism as revealed by the significant nonlinearity
in the stress!strain curves after an initial elastic region.
Yielding is confirmed by unloading tests on bundles
near the failure stress where a large hysteresis and per-
manent deformation is observed in the stress!strain
behavior (Figure 8B inset).
www.acsnano.org
ARTICLE
Figure 9. SEM images detailing subsequent steps (I!III) of
failure of a pristine DWNT bundle. The bundle is found to fail
in a telescopic manner in which inner layers of the bundle
pullout with respect to outer layers. (Top right inset) SEM in-
tensity profiles across the three different bundle layers.
By contrast, heat-treated bundles exhibit almost
purely elastic behavior before failure. Interestingly, in
Figure 8, it is observed that the yield stress estimated
for both heat-treated and pristine bundles is always
greater than the strength of DWNT yarns when consid-
ering a consistent model for calculating stress (the ex-
perimentally determined failure strength of yarns, "1.4
GPa, calculated on the basis of the total cross section
of DWNTs within the bundles, corresponding to a stress
of "5.5 GPa on the outer DWNTs within a bundle). The
implications of this finding with respect to failure mech-
anisms and energy-to-failure will be discussed in the
following section and in more detail in the Supporting
Information.
Two failure mechanisms have been identified in the
bundles with and without polymer. In the case with-
out polymer, the bundles are typically found to exhibit
sword-in-sheath failure in which a single inner bundle
of DWNTs pulls-out after fracture of the outer layers. On
the other hand, bundles with polymer exhibit either
telescopic pullout or complete fracture.
Sword-in-sheath failure, in the case of bundles with-
out polymer, is revealed by high resolution SEM im-
ages of pullout of an individual inner bundle (Figure
8C) for which the difference in diameter of the inner
and outer bundles after pullout ("5 nm) is very close
to 2 times the adjacent axes spacing of DWNTs within
bundles (2 - 2.5 nm), suggesting the failure of a single
outer layer of DWNTs in the bundle. Figure 8E and Fig-
ure 9 show telescopic failure of pristine DWNT bundles.
Here the bundle is found to fail in multiple stages of
pullout where a series of inner bundles slide with re-
spect to the outer DWNTs. This difference in failure
mechanism demonstrates that in the case of DWNT
bundles without heat treatment, upon yielding the
polymer coating introduces stronger lateral interac-
tions between outer DWNTs and DWNTs within a
bundle, leading to either complete fracture or tele-
scopic behavior while carrying load as opposed to full
VOL. 4 ▪ NO. 11 ▪ 6463–6476 ▪ 2010 6471
ARTICLE pullout of all but the outermost layer of DWNTs ob-
served in the heat treated bundle.
DISCUSSION
High strength, high stiffness, and the ability to store
and dissipate a significant amount of energy are among
the most desired properties in advanced fibers. In par-
ticular, a high energy-to-failure in the relatively low
strain regime of approximately 10!20% is of great
interest1,2 for many structural materials as well as in
aerospace and military applications where resistance
to impact and blast loadings is a vital requirement.
However, even to date, concurrent advancements in
all three aspects, namely strength, stiffness, and energy-
to-failure, remain a nearly insurmountable challenge.
This is despite the fact that we have come to realize
some of nature’s design principles in developing hierar-
chical structures with superb mechanical behavior from
weak constituents and the discovery of high
strength!stiffness!toughness building blocks such as
CNTs. Here we have applied a novel design paradigm,
merging structure and materials in the design of a hier-
archical structure, resulting in enhanced mechanical
properties. The yarns presented here benefit from both
long-range interactions between CNT bundles, medi-
ated by polymeric macromolecules, and direct shorter
range CNT!CNT vdW interactions to carry the load and
dissipate energy. The presence of hierarchy in these
yarns and abundance of relatively weak interaction (en-
tropic elasticity of polymers and vdW) resemble the
structure of many biological materials.51
Most commercially available high performance fi-
bers exhibit energy-to-failure in the range of 30!60 J
g!1, while MWNTs, of great interest as individual shells,
can exhibit energy-to-failure of the order of several
thousand J g!1 with failure strains of around 10%.29 To
date, direct spinning of SWNT and DWNT fibers have
been demonstrated to exhibit energy-to-failure values
similar or slightly higher compared to that of commer-
cially available fibers such as carbon fibers and Kevlar.2
All of the DWNT yarns with inherent polymer coatings
studied in this work are found to exhibit values in the
range of 30!100 J g!1 depending on the effective di-
ameter of the yarns, which is comparable to that of the
best available commercial fibers and directly spun CNT
fibers. The energy-to-failure of our thinnest yarns with
an effective diameter of "27 +m is 107 , 17 J g!1,
which is higher than the previously reported values for
similar material systems. In this regard, one may com-
pare these values to the energy-to-failure of CNT-based
yarns fabricated in situ CVD by Motta et al., "60 J g!1,2
to hot-drawn CNT/poly(vinylalcohol) nanocomposite
yarns fabricated by Miaudet et al., ca. 40!60 J g!1,1 and
to CNT yarns spun from a forest of CNTs by Zhang
et al., ca. 14!20 J g!1.3 A comparison between the me-
chanical behavior of CNT yarns obtained from different
fabrication conditions and that of classical reinforce-
6472 VOL. 4 ▪ NO. 11 ▪ NARAGHI ET AL.
ment fibers such as glass, kelvar, and carbon fibers is
presented in a plot of specific failure energy versus spe-
cific strength (Figure 10). A comparison to spider silk is
also included in the figure as a benchmark natural ma-
terial, which exhibits high energy-to-failure in the low to
moderate strain regime. Material systems which dem-
onstrate considerable energy-to-failures only at higher
strains, such as the supertough highly deformable
SWNT system of Miaudet et al.,1 are not discussed in
this section due to their limited applications, as pointed
out earlier in this section. In particular, as shown in the
figure, the improved mechanical performance of our
DWNT-based yarns is owed to superior ductility ("20%
strain) without compromising high strength ("1.4 GPa)
as compared to that shown in refs 1!3. In fact, a quan-
titative mechanistic analysis on the mechanisms of en-
ergy dissipation in pristine and heat-treated yarns (for
details see Supporting Information) reveals that the ma-
jor contribution to the high energy-to-failure in pris-
tine yarns ("100 J g!1), as compared to heat-treated
yarns ("29 J g!1), is a result of interactions between
bundles mediated by polymer coatings molecules. In
addition, for all the yarns studied here, the total energy
stored in DWNTs during mechanical loading is very
small ("12 J g!1). It should be noted that this is far from
approaching the realizable energy-to-failure deter-
mined for individual bundles ("520 J g!1). Moreover,
the stress at failure in the yarns is below the threshold
to induce any permanent deformation in DWNT
bundles.
On the other hand, engineering the interactions be-
tween bundles by the removal of the polymer coating
and the subsequent shrinkage in the yarn diameter,
which promotes vdW interactions, significantly en-
hances the stiffness from 26 , 11 GPa (pristine yarns)
to 67 , 9 GPa (heat-treated yarns) with marginal im-
provement of the strength. However, the elimination
of the long-range interactions between bundles medi-
ated by polymer molecules and replacing them with
the relatively shorter range vdW interactions (compared
to polymeric interactions), results in a loss of ductility
and energy-to-failure by several times.
We conclude the analysis of the energy dissipation
mechanisms by noting the fact that for the small ("2.2
outer diameter) DWNTs in the present study we do not
expect, and have not observed in TEM images, the col-
lapsed tube geometries responsible for high shear in-
teractions reported previously.2 It is interesting to note
from Figure 10 that the heat-treated yarns reported in
this study exhibit similar energy-to-failure as MWNT
yarns spun directly from forests. In both cases we ex-
pect mainly vdW interactions between CNTs and do not
expect collapsed CNT structures given the geometries
of individual tubes within the yarns (10 nm MWNTs3 and
2.2 nm DWNTs [this study]). In contrast, both the poly-
mercoated yarns in this study and in situ CVD grown
CNT yarns with collapsed geometries2 exhibit higher
www.acsnano.org
 ARTICLE
Figure 10. Specific strength as a function of specific energy-to-failure for advanced yarns and fibers.1!8 The upper limit for the ductility
of the materials presented here is taken as ca. 10!20% which roughly corresponds to the ductility of the spider silk at room tempera-
ture, pointing to the high energy-to-failure of a hierarchical material system in nature: (!) In this case, the density of the sample is esti-
mated as the ratio of the area under the stress!strain curve to the reported toughness per unit mass. (!!) The data for glass fibers is ob-
tained from www.AGY.com. Results from refs 2 and 9!11 are reported only for the gage length similar to those studied here (0.8!1.3
cm) to account for the effect of material defects.

energy-to-failure presumably due to higher shear inter-
actions between CNTs. In general, Supporting Informa-
tion Table S1 demonstrates the advantages of stronger
lateral interactions in CNT yarns and also suggests a
more effective improvement through polymer cross-
linking as compared to increased vdW interactions.
Next Generation DWNT Yarn Design. Despite the relatively
high energy-to-failure and strength in our samples, our
multiscale analysis identifies inefficiencies in the utiliza-
tion of the individual DWNT bundles within our system
given that bundles are only loaded to a small fraction of
their strength. The energy analysis presented in the
Supporting Information reveals that the total energy
stored in CNTs is only a small fraction ("2.3%) of the to-
tal energy which can hypothetically be stored in
bundles. This low ratio stems from the relatively weak
lateral interactions between bundles in the current sys-
tem, which does not allow for yarns to reach sufficient
stress at the bundle level to dissipate more energy. In
addition, more elastic energy could be stored in each
bundle if the load was more effectively transferred to in-
ner DWNTs within individual bundles to engage more
DWNTs in the load transfer. Therefore, there are two key
design elements in two scales that need to be ad-
dressed to maximize strength and energy-to-failure in
the CNT-based yarns: (1) enhancing shear interactions
www.acsnano.org
between the bundles such that yarns can sustain higher
loads and increase the dissipated/stored energy in
each bundle (2) enhancing DWNT!DWNT interactions
within each bundle to increase the load bearing cross-
section.
In the current yarn system, failure occurs through
the rupture of the polymer cross-linking molecules (in
the case of pristine yarns) and graphitic type of shear in-
teractions between bundles (in the yarns with no poly-
mer). By utilizing proper polymer cross-linking, in addi-
tion to the inherent polymer coatings of the bundles,
one could in principle substantially enhance the
strength of the bundle!bundle lateral interactions in
the yarns, thus increase the failure strength at the yarn
level, and dissipate/store more energy in each bundle
and in the polymer coating. This goal, for instance,
could be achieved through infiltration of the yarns with
stronger yet resilient polymer cross-links to simulta-
neously improve material strength, stiffness, and
energy-to-failure. The use of polymer crosslinking rep-
resents a more tunable pathway than simply utilizing
vdW interactions, the strength of which is set by the in-
teraction between carbon atoms. Alternatively, one
may induce lateral interactions between bundles by in-
corporating collapsed SWNTs and DWNTs. However,
strain hardening mechanisms in these structures are
VOL. 4 ▪ NO. 11 ▪ 6463–6476 ▪ 2010 6473
ARTICLE primarily limited to CNTs reorientation with the load-
ing direction. Therefore, they are expected to demon-
strate lower strain hardening compared to polymer in-
filtrated yarns. Hence, these structures are prone to
localized failures at relatively low strains as can be con-
firmed by comparing the ductility measured in this
study for pristine yarns with those of both the heat-
treated yarns and the values reported for dogbone CNT
morphologies or by the significant strength size ef-
fects in CVD-grown CNT yarns.2,35
The second design element, enhancing
DWNT!DWNT interactions within each bundle, can be
addressed at the sub-bundle level by the addition of
cross-links between individual DWNTs. Through experi-
ments on individual bundles, we have already demon-
strated that individual bundles with a polymer coating
exhibit a resistance to defects in the form of plastic
yielding after the initial elastic regime, in which addi-
tional energy is dissipated after the failure of the outer
DWNTs through the load transfer to the inner DWNTs
by the polymer coating. Additionally, through particle
radiation, shear interactions between neighboring gra-
phitic layers can be substantially increased to raise the
elastic energy stored within individual bundles. Such an
approach has already been demonstrated to be effec-
tive in cross-linking shells in MWNTs29 and SWNT
bundles.42 This step of cross-linking appears crucial to
the enhancement of the mechanical behavior of the
yarns. In fact, in our elastic modulus measurements of
the bundles, Figure 7, the outer layer of DWNTs in each
bundle carries a substantial portion of the load, leav-
ing inner layers unutilized. In future work these possible
hierarchical material design scenarios can be evalu-
ated by applying multiscale simulation models, through
which we can assess the gain in mechanical properties
by implementing different material treatments and sub-
stantially narrow down the choices for effective design
steps and further material developments.
One last point to mention here is that in the cur-
rent state of the CNT bundle-based yarns, given the
fact that the stress in the yarns even at failure is a small
fraction of the strength of CNTs, the interaction be-
tween the bundles should be considered as the bottle-
EXPERIMENTAL DETAILS
Fabrication of DWNT Yarns. The as-produced mats are highly po-
rous (with a density of "0.04 g cm!3), as shown in Figure 1b,c.
To increase the density and to develop effective load transfer be-
tween CNT bundles, the mats were spun into yarns. For the spin-
ning process, the CNT mats were cut into 1/8, 1/4, and 1/2 inch
wide and several inch long ribbons. Subsequently, they were
twisted on a lathe machine to a few turns per millimeter of
length of the ribbons. The twisting procedure was found to sub-
stantially reduce the fragility of the original mat and to turn the
ribbons into fine yarns with a diameter of ca. 40!100 +m, as
measured in SEM, Figure 1d. The twisting procedure was also ac-
companied by stretching the yarns at an average rate of 2!3
mm min!1 to induce CNT bundle partial alignment. As a conse-
6474 VOL. 4 ▪ NO. 11 ▪ NARAGHI ET AL.
neck which requires improvement to achieve higher
strength!higher toughness materials. This can be real-
ized by comparing the true strength of the yarns (1.4
GPa) with the average strength of the bundles, which
is "6 GPa. The latter is estimated as the ratio of the
maximum force in individual bundle experiments to
the cross section area of all the DWNTs within each
bundle, model A in Figure 7.
CONCLUSION
The mechanical properties of high performance
DWNT!polymer composite yarns, fabricated by twist-
ing and stretching of ribbons of polymer coated
DWNTs, have been measured using a multiscale in situ
SEM experimental approach. The thinnest
DWNT!polymer yarns fabricated were found to ex-
hibit high material strength ("1.4 GPa) as well as
energy-to-failure ("100 J g!1), exceeding that of previ-
ously reported advanced fibers. Through in situ studies,
at nanometer and macroscopic length scales, mecha-
nisms contributing to energy dissipation/storage in the
yarns during axial deformations were identified. Iso-
lated DWNT bundles were found to be capable of stor-
ing/dissipating up to "500 J g!1 at failure. However, the
majority of the energy dissipated at failure in yarns
was attributed to interactions between the polymer
coating and the DWNTs, as unoptimal load transfer be-
tween individual bundles was found to prevent the
stress build up in the yarns required for considerable
energy storage at the bundle level. Two key design ele-
ments to further improve mechanical performance in
the yarns were identified. First, tunable polymer cross-
linking between DWNT bundles was proposed to in-
crease the sustainable load transfer between bundles
and the dissipated/stored energy within each bundle.
Second, particle irradiation was proposed to enhance
DWNT!DWNT interactions within each bundle and in-
crease the load bearing cross-section. By addressing
both these design elements, and applying multiscale
models, we expect to achieve further improvements in
DWNT yarns which already exhibit specific energy-to-
failure superior to the best available synthetic/natural
fibers.
quence of the twisting and drawing of the ribbons, a tension
was developed in the twisted yarn. To avoid premature rupture
and prevent overloading of the yarn during stretching and twist-
ing, a loadcell was used to monitor the applied load on the
yarns. The process of twisting and stretching was continued un-
til a tension force of "80 mN mm!1 per unit width of the ribbon
was generated as measured by the loadcell. Our preliminary
studies on the process of twisting revealed that further stretch-
ing of the yarn resulted in the development of excessive stresses
in the yarn and its failure. During spinning, oleum was sprayed
on the sample which promoted sample relaxation, by allowing
for the mutual sliding of bundles. After spinning, the samples
were left at room temperature for a few hours to dry and re-
form lateral bonds between DWNTs. All DWNT yarns, and
www.acsnano.org

bundles, in our study were fabricated from a DWNT mat with
14.5 wt % of polymer coating as measured using TGA. For this
study, yarns were spun from ribbons with the width of 1/8, 1/4,
and 1/2 in., and the resulting effective diameters of the yarns were
estimated to be 27, 38, and 53 +m, respectively.
Heat Treatment of DWNT Yarns and Bundles. To study the effect of
polymeric coatings on the mechanical behavior of DWNT yarns
and bundles, the coating was removed from the samples by heat
treating them in two steps as follows: (1) Heat the sample in Ar
to 600 °C to remove the polymeric sizing. According to TGA
analysis, Figure 2, this process can remove the sizing almost en-
tirely without damaging the CNTs. (2) Heat the sample in air to
300 °C to burn amorphous carbon which could have formed in
the first step. According to TGA analysis, Figure 2, this process
does not induce defects on the samples. It is to be noted that
Raman spectroscopy confirms that there is no significant change
in the defect density of the DWNTs as a result of the proposed
heat treatment. The ratio of the intensity of the D and G peaks
(ID/IG) in the Raman spectra before and after heat treatment is
similar and very low (ca. 0.02!0.03), where a marginal decrease
is found for the heat-treated samples suggesting that a small
number of defects are annealed out. The low ID/IG ratio in both
cases indicates that the DWNTs are of very high quality, nearly
void of structural defects.52 In the case of yarns, the process of
heat treatment was performed after they were spun, since the re-
moval of the polymers from the mats made them extremely frag-
ile and very difficult to spin. On the other hand, for the nano-
mechanical experiments, the heat-treated DWNT bundles were
extracted from heat-treated mats.
In Situ SEM Mechanical Testing of Yarns. A microtensile tester was
used, which was adapted to function inside the SEM. The stage
was actuated using a linear motor at a speed of "16 +m sec!1,
which corresponds to a strain rate of 1.6 - 10!3 s!1 for a typical
gage length of 1 cm. The stage was equipped with a miniature
loadcell which allowed for the measurement of tensile loads as
high as 2.5 N, with a resolution of as low as 2 mN. The crosshead
deformation of the sample was recorded using a linear variable
differential transformer (LVDT) with a resolution of 1 +m (strain
resolution of "0.01%), which was used to calculate the average
strain in the sample. All the components of the stage, the load-
cell, motor, and LVDT, were vacuum compatible. Samples were
gripped to the testing device using a resin-based epoxy. No
sample sliding was observed during the in situ SEM experiments.
This is consistent with the fact that, prior to sample failure in
the gage sections, no sudden drops in load, which would have
been suggestive of sample sliding in the grips, were observed.
In Situ SEM Mechanical Testing of Isolated Bundles. A combined
nanomanipulator/cantilever method was applied to test indi-
vidual bundles of DWNTs which have been isolated from the as-
fabricated mat material. Individual DWNT bundles, with a diam-
eter ranging between 30 and 55 nm, were first isolated onto Cu
TEM grids by lightly rubbing the grids over the bulk DWNT mat
material. For each test a bundle was attached between a tung-
sten tip and the integrated tip of a single crystal Si cantilever us-
ing a nanomanipulator (Klocke). Manipulation was aided by elec-
tron beam etching of the DWNT bundles during the
manipulation procedure.53 The etching procedure also allowed
control over the length of tested bundles. Clamping between the
bundle and tips was achieved by electron beam induced depo-
sition (EBID) of Pt.54 Tensile tests were performed by the con-
trolled displacement of the tungsten nanomanipulator tip (see
Figure 6). The force acting on the bundle was measured using
high resolution SEM images of the deflection of the Si cantile-
ver, for which normal force constants were calibrated for each
beam using a standard geometrical calibration method.55 For
each tensile test the final position of the Si tip after failure of the
bundle was used to correct for any drift during testing.
Acknowledgment. H. D. Espinosa gratefully acknowledges
support from ONR through award No. N00014-08-1-0108 and
the ARO through award No. W911NF-09-1-0541. The authors
gratefully acknowledge fruitful discussions with Dr. I. Perez. The
authors also acknowledge Dr. S. Li of the Electron Probe Instru-
mentation Center (EPIC) at Northwestern University for obtaining
www.acsnano.org
ARTICLE
TEM images and Dr. O. Compton for collection of Raman data
on DWNT mats.
Supporting Information Available: The calculation of the den-
sity of CNT bundles in a hexagonally closed packed structure; es-
timation of the number of DWNTs carrying the load used for
nanomechanical experiment on individual bundles of DWNTs;
proposed method and calculations for the quantification of the
mechanisms of energy dissipation and storage in DWNT yarns.
This material is available free of charge via the Internet at http://
pubs.acs.org.
REFERENCES AND NOTES
1. Miaudet, P.; Badaire, S.; Maugey, M.; Derre, A.; Pichot, V.;
Launois, P.; Poulin, P.; Zakri, C. Hot-Drawing of Single and
Multiwall Carbon Nanotube Fibers for High Toughness
and Alignment. Nano Lett. 2005, 5, 2212–2215.
2. Motta, M.; Moisala, A.; Kinloch, I.; Windle, A. High
Performance Fibres from Dog Bone Carbon Nanotubes.
Adv. Mater. 2007, 19, 3721–3726.
3. Zhang, M.; Atkinson, K. R.; Baughman, R. H. Multifunctional
Carbon Nanotube Yarns by Downsizing an Ancient
Technology. Science 2004, 306, 1358–1361.
4. Dalton, A. B.; Collins, S.; Munoz, E.; Razal, J. M.; Ebron, V. H.;
Ferraris, J. P.; Coleman, J. N.; Kim, B. G.; Baughman, R. H.
Super-Tough Carbon-Nanotube FibresOThese
Extraordinary Composite Fibres Can Be Woven into
Electronic Textiles. Nature 2003, 423, 703–706.
5. Vollrath, F.; Knight, D. P. Liquid Crystalline Spinning of
Spider Silk. Nature 2001, 410, 541–548.
6. Vigolo, B.; Penicaud, A.; Coulon, C.; Sauder, C.; Pailler, R.;
Journet, C.; Bernier, P.; Poulin, P. Macroscopic Fibers and
Ribbons of Oriented Carbon Nanotubes. Science 2000,
290, 1331–1334.
7. Kromm, F. X.; Lorriot, T.; Coutand, B.; Harry, R.; Quenisset,
J. M. Tensile and Creep Properties of Ultrahigh Molecular
Weight PE Fibres. Polym. Test. 2003, 22, 463–470.
8. Rath, J. P.; Chaki, T. K.; Khastgir, D. Change in Fiber
Properties Due to the Heat Treatment of Nylon 6 Tire
Cords. J. Appl. Polym. Sci. 2008, 108, 3960–3967.
9. Koziol, K.; Vilatela, J.; Moisala, A.; Motta, M.; Cunniff, P.;
Sennett, M.; Windle, A. High-Performance Carbon
Nanotube Fiber. Science 2007, 318, 1892–1895.
10. Motta, M.; Li, Y. L.; Kinloch, I.; Windle, A. Mechanical
Properties of Continuously Spun Fibers of Carbon
Nanotubes. Nano Lett. 2005, 5, 1529–1533.
11. Stano, K.; Koziol, K.; Pick, M.; Motta, M.; Moisala, A.; Vilatela,
J.; Frasier, S.; Windle, A. Direct Spinning of Carbon
Nanotube Fibres from Liquid Feedstock. Int. J. Mater. Form.
2008, 1, 59–62.
12. Gupta, H. S.; Seto, J.; Wagermaier, W.; Zaslansky, P.;
Boesecke, P.; Fratzl, P. Cooperative Deformation of Mineral
and Collagen in Bone at the Nanoscale. Proc. Natl. Acad.
Sci. U.S.A. 2006, 103, 17741–17746.
13. Barthelat, F.; Tang, H.; Zavattieri, P. D.; Li, C. M.; Espinosa,
H. D. On the Mechanics of Mother-of-Pearl: A Key Feature
in the Material Hierarchical Structure. J. Mech. Phys. Solids
2007, 55, 306–337.
14. Jackson, A. P.; Vincent, J. F. V.; Turner, R. M. The
Mechanical Design of Nacre. Proc. R. Soc. London 1988,
234, 415–440.
15. Qin, Z.; Kreplak, L.; Buehler, M. J. Hierarchical Structure
Controls Nanomechanical Properties of Vimentin
Intermediate Filaments. PLoS One 2009, 4, e7294.
16. Buehler, M. J.; Yung, Y. C. How Protein Materials Balance
Strength, Robustness and Adaptability. HFSP J. 2010, 4,
26–40.
17. Lakes, R. Materials with Structural Hierarchy. Nature 1993,
361, 511–515.
18. Espinosa, H.; Rim, J.; Barthelat, F.; Buehler, M. J. Merger of
Structure and Material in Nacre and Bone-Perspectives on
de Novo Biomimetic Materials. Prog. Mater. Sci. 2009, 54,
1059–1100.
19. Ritchie, R. O.; Buehler, M. J.; Hansma, P. Plasticity and
Toughness in Bone. Phys. Today 2009, 62, 41–47.
VOL. 4 ▪ NO. 11 ▪ 6463–6476 ▪ 2010 6475
ARTICLE 20. Qin, Z.; Buehler, M. J.; Kreplak, L. A Multiscale Approach to
Understand the Mechanobiology of Intermediate
40. Yu, M. F.; Files, B.; Arepalli, S.; Ruoff, R. Tensile Loading of
Ropes of Single Wall Carbon Nanotubes and their
Filaments. J. Biomech. 2010, 43, 15–22. Mechanical Properties. Phys. Rev. Lett. 2000, 84,
21. Gao, H. J.; Ji, B. H.; Jager, I. L.; Arzt, E.; Fratzl, P. Materials 5552–5555.
Become Insensitive to Flaws at Nanoscale: Lessons from 41. Yu, M. F.; Lourie, O.; Dyer, M. J.; Moloni, K.; Kelly, T. F.;
Nature. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 5597–5600. Ruoff, R. S. Strength and Breaking Mechanism of
22. Mayer, G.; Sarikaya, M. Rigid Biological Composite Multiwalled Carbon Nanotubes under Tensile Load.
Materials: Structural Examples for Biomimetic Design. Exp. Science 2000, 287, 637–640.
Mech. 2002, 42, 395–403. 42. Kis, A.; Csanyi, G.; Salvetat, J. P.; Lee, T. N.; Couteau, E.;
23. Wang, C. A.; Huang, Y.; Zan, Q. F.; Guo, H.; Cai, S. Y. Kulik, A. J.; Benoit, W.; Brugger, J.; Forro, L. Reinforcement
Biomimetic Structure DesignOA Possible Approach To of Single-Walled Carbon Nanotube Bundles by Intertube
Change the Brittleness of Ceramics in Nature. Mater. Sci. Bridging. Nat. Mater. 2004, 3, 153–157.
Eng. C 2000, 11, 9–12. 43. Jin, H.-Z.; He, R.-R.; Zhu, J. Helicity and Inter-tube Bonding
24. Verma, D.; Katti, K.; Katti, D.; Mohanty, B. Mechanical in Bundles of Single-Walled Carbon Nanotubes. J. Electron
Response and Multilevel Structure of Biomimetic Microsc. 1999, 48, 339–343.
Hydroxyapatite/Polygalacturonic/Chitosan 44. Lima, A. M. F.; Musumeci, A. W.; Liu, H. W.; Waclawik, E. R.;
Nanocomposites. Mater. Sci. Eng. C 2008, 28, 399–405. Silva, G. G. Purity Evaluation and Influence of Carbon
25. Meyers, M. A.; Lin, A. Y. M.; Seki, Y.; Chen, P. Y.; Kad, B. K.; Nanotube on Carbon Nanotube/Graphite Thermal
Bodde, S. Structural Biological Composites: An Overview. Stability. J. Therm. Anal. Calorim. 2009, 97, 257–263.
JOM 2006, 58, 35–41. 45. Pang, L. S. K.; Saxby, J. D.; Chatfield, S. P.
26. Espinosa, H. D.; Juster, A. L.; Latourte, F. J.; Loh, O. Y.; Thermogravimetric Analysis of Carbon Nanotubes and
Gregoire, D.; Zavattieri, P. D. Lessons in Abalone Shell Nanoparticles. J. Phys. Chem. 1993, 97, 6941–6942.
Toughness Applied to Synthetic Materials. Unpublished 46. Jones, N. Elastic-Plastic and Viscoelastic Behavior of a
work. Continuous Filament Yarn. Int. J. Mech. Sci. 1974, 16, 679–
27. Yang, Y.; Chen, X.; Shao, Z.; Zhou, P.; Porter, D.; Knight, 687.
D. P.; Vollrath, F. Toughness of Spider Silk at High and Low 47. Hamm, E.; Roman, B.; Melo, F. Dynamics of Developable
Temperatures. Adv. Mater. 2005, 17, 84–88. Cones under Shear. Phys. Rev. E 2004, 70, 026607.
28. Locascio, M.; Peng, B.; Zapol, P.; Zhu, Y.; Li, S.; Belytschko, 48. Boudaoud, A.; Patricio, P.; Couder, Y.; Ben Amar, M.
T.; Espinosa, H. D. Tailoring the Load Carrying Capacity of Dynamics of Singularities in a Constrained Elastic Plate.
MWCNTs Through Inter-shell Atomic Bridging. Exp. Mech. Nature 2000, 407, 718–720.
2009, 49, 169–182. 49. Leech, C. M. The Modelling of Friction in Polymer Fibre
29. Peng, B.; Locascio, M.; Zapol, P.; Li, S.; Mielke, S. L.; Schatz, Ropes. Int. J. Mech. Sci. 2002, 44, 621–643.
G. C.; Espinosa, H. D. Measurements of Near-Ultimate 50. Qian, D.; Liu, W. K.; Ruoff, R. S. Load Transfer Mechanism in
Strength for Multiwalled Carbon Nanotubes and Carbon Nanotube Ropes. Compos. Sci. Technol. 2003, 63,
Irradiation-Induced Crosslinking Improvements. Nat. 1561–1569.
51. Buehler, M. J.; Ackbarow, T. Fracture Mechanics of Protein
Nanotechnol. 2008, 3, 626–631.
Materials. Mater. Today 2007, 10, 46–58.
30. Haskins, R. W.; Maier, R. S.; Ebeling, R. M.; Marsh, C. P.;
52. Dresselhaus, M. S.; Dresselhaus, G.; Saito, R.; Jorio, A.
Majure, D. L.; Bednar, A. J.; Welch, C. R.; Barker, B. C. Tight-
Raman Spectroscopy of Carbon Nanotubes. Phys. Rep.
Binding Molecular Dynamics Study of the Role of Defects
2005, 409, 47–99.
on Carbon Nanotube Moduli and Failure. J. Chem. Phys.
53. Yuzvinsky, T. D.; Fennimore, A. M.; Mickelson, W.; Esquivias,
2007, 127, 074708.
C.; Zettl, A. Precision Cutting of Nanotubes with a Low-
31. Pop, E.; Mann, D. A.; Goodson, K. E.; Dai, H. J. Electrical and
Energy Electron Beam. Appl. Phys. Lett. 2005, 86, 05310-9!
Thermal Transport in Metallic Single-Wall Carbon
05310-3.
Nanotubes on Insulating Substrates. J. Appl. Phys. 2007,
54. Agrawal, R.; Peng, B.; Gdoutos, E. E.; Espinosa, H. D.
101, 093710.
Elasticity Size Effects in ZnO NanowiresOA Combined
32. Collins, P.; Arnold, M.; Avouris, P. Engineering Carbon
Experimental!Computational Approach. Nano Lett. 2008,
Nanotubes and Nanotube Circuits Using Electrical 8, 3668–3674.
Breakdown. Science 2001, 292, 706–709. 55. Meyer, E.; Hug, H.; Bennewitz, R. Scanning Probe
33. Pop, E.; Mann, D.; Wang, Q.; Goodson, K.; Dai, H. Thermal Microscopy: The Lab on a Tip; Springer-Verlag: Berlin, 2004.
Conductance of an Individual Single-wall Carbon
Nanotube above Room Temperature. Nano Lett. 2006, 6,
96–100.
34. Li, Y.-L.; Kinloch, I.; Windle, A. Direct Spinning of Carbon
Nanotube Fibers from Chemical Vapor Deposition
Synthesis. Science 2004, 304, 276–278.
35. Stano, K. L.; Koziol, K.; Pick, M.; Motta, M. S.; Moisala, A.;
Vilatela, J. J.; Frasier, S.; Windle, A. H. Direct Spinning of
Carbon Nanotube Fibres from Liquid Feedstock. Int. J.
Mater. Form. 2008, 1, 4.
36. Elliott, J.; Sandler, J.; Windle, A.; Young, R.; Shaffer, M.
Collapse of Single-Wall Carbon Nanotubes Is Diameter
Dependent. Phys. Rev. Lett. 2004, 92, 95501.
37. Dalton, A. B.; Collins, S.; Munoz, E.; Razal, J. M.; Ebron, V. H.;
Ferraris, J. P.; Coleman, J. N.; Kim, B. G.; Baughman, R. H.
Super-tough Carbon-Nanotube FibresOThese
Extraordinary Composite Fibres Can Be Woven into
Electronic Textiles. Nature 2003, 423, 703.
38. Zhang, M.; Fang, S. L.; Zakhidov, A. A.; Lee, S. B.; Aliev, A. E.;
Williams, C. D.; Atkinson, K. R.; Baughman, R. H. Strong,
Transparent, Multifunctional, Carbon Nanotube Sheets.
Science 2005, 309, 1215–1219.
39. Liu, K.; Sun, Y. H.; Zhou, R. F.; Zhu, H. Y.; Wang, J. P.; Liu, L.;
Fan, S. S.; Jiang, K. L. Carbon Nanotube Yarns with High
Tensile Strength Made by a Twisting and Shrinking
Method. Nanotechnology 2010, 21, 045708.

6476 VOL. 4 ▪ NO. 11 ▪ NARAGHI ET AL. www.acsnano.org